|
--- |
|
library_name: transformers |
|
license: apache-2.0 |
|
base_model: openai/whisper-tiny |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- wer |
|
model-index: |
|
- name: whisper-tiny-luganda-final |
|
results: [] |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
# whisper-tiny-luganda-final |
|
|
|
This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.4549 |
|
- Wer: 0.4660 |
|
|
|
## Model description |
|
|
|
More information needed |
|
|
|
## Intended uses & limitations |
|
|
|
More information needed |
|
|
|
## Training and evaluation data |
|
|
|
More information needed |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 5e-05 |
|
- train_batch_size: 8 |
|
- eval_batch_size: 4 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 2 |
|
- total_train_batch_size: 16 |
|
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments |
|
- lr_scheduler_type: linear |
|
- lr_scheduler_warmup_ratio: 0.05 |
|
- training_steps: 21000 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Wer | |
|
|:-------------:|:------:|:-----:|:---------------:|:------:| |
|
| 2.0498 | 0.1088 | 500 | 0.9947 | 0.8351 | |
|
| 1.5134 | 0.2175 | 1000 | 0.7854 | 0.7198 | |
|
| 1.2494 | 0.3263 | 1500 | 0.6915 | 0.7817 | |
|
| 1.1587 | 0.4351 | 2000 | 0.6377 | 0.7916 | |
|
| 1.0453 | 0.5438 | 2500 | 0.5958 | 0.7276 | |
|
| 1.0065 | 0.6526 | 3000 | 0.5708 | 0.5803 | |
|
| 0.935 | 0.7614 | 3500 | 0.5478 | 0.5882 | |
|
| 0.8838 | 0.8701 | 4000 | 0.5256 | 0.5876 | |
|
| 0.8956 | 0.9789 | 4500 | 0.5123 | 0.5380 | |
|
| 0.6998 | 1.0877 | 5000 | 0.5078 | 0.5332 | |
|
| 0.6735 | 1.1964 | 5500 | 0.4976 | 0.5400 | |
|
| 0.7159 | 1.3052 | 6000 | 0.4934 | 0.5097 | |
|
| 0.6693 | 1.4140 | 6500 | 0.4843 | 0.5043 | |
|
| 0.6513 | 1.5227 | 7000 | 0.4774 | 0.5038 | |
|
| 0.6478 | 1.6315 | 7500 | 0.4736 | 0.5015 | |
|
| 0.6554 | 1.7403 | 8000 | 0.4634 | 0.5042 | |
|
| 0.6491 | 1.8490 | 8500 | 0.4608 | 0.4941 | |
|
| 0.6636 | 1.9578 | 9000 | 0.4526 | 0.4774 | |
|
| 0.4392 | 2.0666 | 9500 | 0.4581 | 0.4788 | |
|
| 0.4567 | 2.1753 | 10000 | 0.4575 | 0.4842 | |
|
| 0.4383 | 2.2841 | 10500 | 0.4562 | 0.4787 | |
|
| 0.4479 | 2.3929 | 11000 | 0.4546 | 0.4747 | |
|
| 0.4431 | 2.5016 | 11500 | 0.4517 | 0.4820 | |
|
| 0.4354 | 2.6104 | 12000 | 0.4498 | 0.4612 | |
|
| 0.4956 | 2.7192 | 12500 | 0.4442 | 0.4825 | |
|
| 0.4427 | 2.8279 | 13000 | 0.4454 | 0.4693 | |
|
| 0.4371 | 2.9367 | 13500 | 0.4430 | 0.4573 | |
|
| 0.2794 | 3.0455 | 14000 | 0.4429 | 0.4544 | |
|
| 0.2826 | 3.1542 | 14500 | 0.4491 | 0.4694 | |
|
| 0.2914 | 3.2630 | 15000 | 0.4497 | 0.4536 | |
|
| 0.3065 | 3.3718 | 15500 | 0.4501 | 0.4557 | |
|
| 0.2879 | 3.4805 | 16000 | 0.4492 | 0.4532 | |
|
| 0.2703 | 3.5893 | 16500 | 0.4495 | 0.4535 | |
|
| 0.269 | 3.6981 | 17000 | 0.4466 | 0.4681 | |
|
| 0.2834 | 3.8068 | 17500 | 0.4445 | 0.4686 | |
|
| 0.2758 | 3.9156 | 18000 | 0.4470 | 0.4660 | |
|
| 0.1819 | 4.0244 | 18500 | 0.4497 | 0.4640 | |
|
| 0.1855 | 4.1331 | 19000 | 0.4510 | 0.4611 | |
|
| 0.1832 | 4.2419 | 19500 | 0.4543 | 0.4693 | |
|
| 0.1826 | 4.3507 | 20000 | 0.4558 | 0.4640 | |
|
| 0.1779 | 4.4594 | 20500 | 0.4553 | 0.4672 | |
|
| 0.1821 | 4.5682 | 21000 | 0.4549 | 0.4660 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.47.1 |
|
- Pytorch 2.5.1+cu124 |
|
- Datasets 3.2.0 |
|
- Tokenizers 0.21.0 |
|
|
